Kwimp/speed_augmentation

1{2  "activation_dropout": 0.0,3  "activation_function": "gelu",4  "apply_spec_augment": false,5  "architectures": [6    "WhisperForConditionalGeneration"7  ],8  "attention_dropout": 0.0,9  "begin_suppress_tokens": [10    220,11    5025712  ],13  "bos_token_id": 50257,14  "classifier_proj_size": 256,15  "d_model": 768,16  "decoder_attention_heads": 12,17  "decoder_ffn_dim": 3072,18  "decoder_layerdrop": 0.0,19  "decoder_layers": 12,20  "decoder_start_token_id": 50258,21  "dropout": 0.0,22  "dtype": "float32",23  "encoder_attention_heads": 12,24  "encoder_ffn_dim": 3072,25  "encoder_layerdrop": 0.0,26  "encoder_layers": 12,27  "eos_token_id": 50257,28  "forced_decoder_ids": null,29  "init_std": 0.02,30  "is_encoder_decoder": true,31  "mask_feature_length": 10,32  "mask_feature_min_masks": 0,33  "mask_feature_prob": 0.0,34  "mask_time_length": 10,35  "mask_time_min_masks": 2,36  "mask_time_prob": 0.05,37  "max_source_positions": 1500,38  "max_target_positions": 448,39  "median_filter_width": 7,40  "model_type": "whisper",41  "num_mel_bins": 80,42  "pad_token_id": 50257,43  "scale_embedding": false,44  "suppress_tokens": [],45  "tie_word_embeddings": true,46  "transformers_version": "5.8.1",47  "use_cache": false,48  "use_weighted_layer_sum": false,49  "vocab_size": 5186550}51